[OE-core] psplash activation state w/ systemd

[Alex Kiernan]

On Wed, Mar 11, 2020 at 5:27 PM Alex Kiernan <alex.kiernan at gmail.com> wrote:
>
> On Wed, Mar 11, 2020 at 5:20 PM Richard Purdie
> <richard.purdie at linuxfoundation.org> wrote:
> >
> > On Mon, 2020-03-09 at 15:18 +0000, Alex Kiernan wrote:
> > > I've a branch with systemd 245 on it which fails testing because
> > > psplash gets restarted all the time.
> > >
> > > But ignoring the systemd 245 piece, it looks to me like psplash could
> > > be restarted under systemd 244 too as the main process exits and our
> > > units aren't marked RemainAfterExit. I haven't figured out what's
> > > triggering it in 245 and not 244, but I suspect it's something
> > > similar
> > > to this:
> > >
> > > https://github.com/systemd/systemd/commit/9fd32ff7d363945fbf8fdae0128702b995127558
> > >
> > > Given where we are in the release cycle and how painful psplash has
> > > obviously been, I'm inclined to leave it until after dunfell ships
> > > and then do it as part of the whole 245 upgrade?
> > >
> > > Equally if you see psplash flapping during tests, it's almost
> > > certainly this problem.
> >
> > Thanks for the report. The weird thing is that master is working with
> > this as far as I can see, certainly no test failures since the last fix
> > I pushed a few days ago.
> >
> > Is there some specific change to the unit files we should make? I'm a
> > little reluctant to pull this out now given we feel like we're close to
> > it working.
> >
>
> It's basically trivial (add RemainAfterExit=yes to both units), but as
> you say, given where we are and it appears to work, I don't think
> now's the time to do it.
>
> I suspect something's changed in 245 which makes it far more likely
> that you see units retriggered, but I can't for the life of me work
> out which commit it is (I should stop trying to do it by inspection
> and bisect it out...).
>

Looks like it's:

https://github.com/systemd/systemd/commit/097537f07a2fab3cb73aef7bc59f2a66aa93f533

which I found via:

https://github.com/systemd/systemd/issues/15091

Definitely ignore this issue for now... I suspect we do actually want
the RemainAfterExit, but for now wait and watch where upstream systemd
ends up on this seems like the sensible approach.

-- 
Alex Kiernan